Set up your environment totake the TensorFlowDeveloper Ce�i�cate Exam

Questions? Email tenso�low-ce�i�cate-suppo�@google.com.

Last Updated: December 14, 2020

This document describes how to get your environment ready to take the TensorFlowDeveloper Ce�i�cate exam. It does not discuss what the exam is or how to take it. Please go totenso�low.org/ce�i�cate for more information about the TensorFlow Developer Ce�i�cateprogram.

Contents

Before you begin

Refund policy

Install Python 3.8

Install PyCharm

Get your environment ready for the examWhat libraries will the exam infrastructure install?Make sure that PyCharm isn't subject to �le-loading controlsWindows and Linux Users: Check your GPU driversMac Users: Ensure you have Python 3.8All Users: Create a Test Vi�ual Environment that uses TensorFlow in PyCharm

Create a new PyCharm projectInstall TensorFlow and related packagesCheck the suppo�ed version of Python in PyCharmPractice training TensorFlow models

Setup your environment for the TensorFlow Developer Certificate exam 1

FAQsHow do I sta� the exam?What version of TensorFlow is used in the exam?Is there a minimum hardware requirement?Where is the candidate handbook?

Before you beginThe TensorFlow ce�i�cate exam runs inside PyCharm.

You need to use TensorFlow 2.3.

You must use Python 3.8 to ensure accurate grading of your models. The exam has beentested with Python 3.8.0 and TensorFlow 2.3.0.

Before you sta� the exam, make sure your environment is ready:

❏ Make sure you have Python 3.8 installed on your computer.

❏ Check that your system meets the installation requirements for PyCharm here.

❏ Install the latest stable version of PyCharm.

❏ Ensure that you have a stable internet connection before sta�ing the exam.

❏ Make sure you are familiar with using PyCharm and that it is con�gured correctly.

See the section Before you begin: Get your environment ready fu�her down thisdocument for details on making sure that you have PyCharm ready and set up correctlyfor the exam.

Setup your environment for the TensorFlow Developer Certificate exam 2

Refund policyWhen the exam sta�s, you will have �ve hours to complete it.

It is your responsibility to ensure that

● you are using a computing environment that can run PyCharm and PyCharm pluginswithout problems.

● you have a strong and reliable internet connection for the duration of the exam.

● you are using a computing environment that can handle the creation, training andtesting of machine learning models. See Is there a minimum hardware requirement?

● you carefully read the instructions that will be sent to you when you purchase the exam.

● you read and understand the Terms and Conditions before you take the exam.

Impo�ant! A�er you sta� the exam, we are unable to issue refunds orretakes for any reason.

To sta� the exam, go to tenso�low.org/ce�i�cate.

Install Python 3.8IMPORTANT! Make sure Python 3.8 is installed on your machine before you sta� the exam.Although TensorFlow works with other versions of Python, the exam has been tested withPython 3.8.0.

For step by step instructions to updating Python on your workstation, see Before you begin:Get your environment ready to take the exam (fu�her down in this document).

Don't use Anaconda.

IMPORTANT! Your Python environment must be installed from canonical Python packagesavailable at python.org. We cannot be responsible for errors in your submissions due toself-built or non-canonical Python environments.

Setup your environment for the TensorFlow Developer Certificate exam 3

Install PyCharm1. Check that your system meets the installation requirements for PyCharm here.

2. Install the latest stable version of PyCharm 2020.2 (the exam doesn't yet work in2020.3):

○ You can use either PyCharm Professional or PyCharm Community Edition.

○ Don't use the Anaconda plugin..

Get your environment ready for theexamYou will take the exam inside PyCharm. The exam infrastructure will create a project and installsome so�ware packages for you. You can install whatever other so�ware you need as you takethe exam. It is impo�ant that you don't change the version of TensorFlow within the examproject, because the grading infrastructure uses the same version of TensorFlow as the exam.

What libraries will the exam infrastructure install?

When you sta� the TensorFlow developer ce�i�cate exam, the exam framework will install thefollowing so�ware in your PyCharm project:

tensorflow==2.3.0

tensorflow-datasets==3.2.1

Pillow==7.2.0

Setup your environment for the TensorFlow Developer Certificate exam 4

Before you begin the exam, to mitigate any possible issues that you could encounter with theexam environment, we recommend creating a PyCharm project, loading the same librariesthat you will need in the exam, and practicing training models in PyCharm.

Make sure that PyCharm isn't subject to �le-loading controls

The TensorFlow Developer ce�i�cate exam needs to load �les. Make sure that the so�ware orcomputer that you will use to access PyCharm isn't subject to any controls that could block theloading of �les. This advice is relevant, for example, if you plan to use a Windows serveradministered by your company which might have restrictions on what �les can be loadedthrough the server.

Windows and Linux Users: Check your GPU drivers

If you are a Windows user, and you have NVidia graphics cards installed, make sure your driverswork with the required version of TensorFlow. We recommend you test out a system a�erfollowing the steps here: tenso�low.org/install/gpu. See the instructions below for testing in aPyCharm vi�ual environment.

Mac Users: Ensure you have Python 3.8

The exam has been tested with Python 3.8. Macs ship with 2.7 as the default native interpreter,and PyCharm can get confused by this. Merely installing Python 3.8 is not enough. First ensurethat Python 3.8 is properly installed, by typing ‘python3’ at the terminal. This command returnsthe version of Python 3 that is installed – makes sure it says Python 3.8.x.

If it isn’t installed, go to python.org/downloads to download a version of Python 3.8. Note thatthis is not the latest version. If you use homebrew make sure that you use it to install Python3.8, and not the latest (which it will default to).

All Users: Create a Test Vi�ual Environment that uses TensorFlow inPyCharm

When you sta� the TensorFlow developer ce�i�cate exam, it will install TensorFlow for you.However, we strongly recommend that you practice using TensorFlow in PyCharm before you

Setup your environment for the TensorFlow Developer Certificate exam 5

sta� the exam, so that you are familiar with the environment that you will work in while takingthe exam, and also to mitigate any con�guration issues.

There are many varied computers with di�erent operating systems, GPUs, and otherdependencies. It’s strongly recommended that you follow these steps to ensure that youhave a well-con�gured PyCharm environment that TensorFlow works on before sta�ing theexam.

This will help you ensure that you have the right Python environment, that your GPU driverswork correctly, and that you are able to install the required dependencies.

Create a new PyCharm project

Open PyCharm and sta� a new project in PyCharm. Call it TFExamTest (or whatever you like).

When asked for the Project Interpreter, select ‘New Vi�ualenv Environment’ as shown here.Make sure that the Base Interpreter is Python 3.8.

Setup your environment for the TensorFlow Developer Certificate exam 6

If it is a Python 3.8 (as shown in the screenshot above), click Create. If it isn’t, exit PyCharm,and install Python 3.8 on your machine and then try this step again. In pa�icular, if youare a Mac user check the section above.

Install TensorFlow and related packages

If you were able to create the environment, open the PyCharm preferences dialog. (Windowsand Linux: File->Se�ings. On Mac it is PyCharm->Preferences)

Choose the Project, and then Project Interpreter as shown here:

Setup your environment for the TensorFlow Developer Certificate exam 7

Press the + bu�on. It’s at the bo�om of the screenshot above, but in some cases it might be onthe right hand side. This will open the ‘Available Packages’ window, with details on packagesthat you can install.

Search for "tenso�low. Note that the exam uses a speci�c version, so change the version (ifappropriate) if you want to emulate the exam environment.

Setup your environment for the TensorFlow Developer Certificate exam 8

To install TensorFlow, click the ‘Install Package’ link at the bo�om. Repeat this process for thefollowing packages:

● tenso�low

● tenso�low-datasets

● numpy

● Pillow

● urllib3

You will now have a working environment with all the dependencies required for the exam. Youcan test the environment with some simple Python and TensorFlow code to be sure.

If you are a Windows or Linux GPU user, a simple Python script you can try is:

import tensorflow as tfprint(tf.__version__)

If your drivers are con�gured correctly, this script will execute. If not, check your driverdependencies, and the dependencies listed at: tenso�low.org/install

Check the suppo�ed version of Python in PyCharm

If you aren’t sure what version of Python is being used in your PyCharm environment, go thethe ‘Python Console’ window in PyCharm, which will usually be a tab at the bo�om of yourscreen, and type

print(sys.version)

The terminal will repo� back the Python version as shown here:

PyDev console: starting.

>>>sys.version'3.8.6 (v3.8.6:db455296be, Sep 23 2020, 13:31:39)

Setup your environment for the TensorFlow Developer Certificate exam 9

Practice training TensorFlow models

The TensorFlow Developer Ce�i�cate exam will require you to create and train models similarto those found on tenso�low.org/tutorials, such as:

● Image classi�cation

● Word embeddings

We recommend that you practice running the code from these tutorials in PyCharm tobecome familiar with using PyCharm to train TensorFlow models.

FAQs

How do I sta� the exam?

This document does not cover how to take the exam. Please go to tenso�low.org/ce�i�catefor more information about the TensorFlow Developer Ce�i�cate program and for links to sta�the exam.

What version of TensorFlow is used in the exam?

The exam will test you on your ability to de�ne, train and test models with TensorFlow 2.3.0.

Regardless of what TensorFlow version you install on your machine, the exam infrastructurerequires, and will install, the version listed in What libraries will the exam infrastructure install?

Is there a minimum hardware requirement?

We do not specify a minimum hardware requirement. We rely on you, as an experienced MLpractitioner, to know the capabilities of your own environment with regard to training MLmodels with TensorFlow.

The TensorFlow Developer Ce�i�cate exam will require you to create and train models similarto those found on tenso�low.org/tutorials (such as Example 1, Example 2). We recommendthat you practice these to make sure your hardware can handle them.

Setup your environment for the TensorFlow Developer Certificate exam 10

You will take the exam inside PyCharm. You can get your environment ready following theinstructions in this document and then run your practice models in PyCharm. You can do thisbefore you purchase the exam.

We allow 5 hours for the exam because we know that it will take some time to train the models.

During the exam, you are welcome to experiment with training models using GCP, AWS,Jupyter Notebooks or Google Colab, but you will still need to de�ne, train and save yourmodels within the exam environment, inside PyCharm, in order to submit it for grading.

How do I set up GPU and CUDA?

If you use GPU and CUDA, see the o�cial TensorFlow installation pages:

● h�ps://www.tenso�low.org/install● h�ps://www.tenso�low.org/install/gpu

See the section GPU: So�ware Requirements for information on what version of CUDA is usedby TensorFlow. Regardless of what TensorFlow version you install on your machine, the examinfrastructure requires, and will install, the version listed in What libraries will the examinfrastructure install? So make sure that you have installed the appropriate version of CUDA (Ifyou use it) for the version of TensorFlow that the exam uses.

Where is the candidate handbook?

The candidate handbook is linked from tenso�low.org/ce�i�cate.

Setup your environment for the TensorFlow Developer Certificate exam 11